• Carvajal-Castrillón, Alejandro
• Alvarez-Montoya, Joham
• Sierra, Julian

Abstract

<p>Aiming to provide more efficient, lightweight structures, composite materials are being extensively used in aerospace vehicles. As the failure mechanisms of these materials are complex, damage detection becomes challenging, requiring advanced techniques for assessing structural integrity and maintaining aircraft safety. In this context, Structural Health Monitoring (SHM) seeks for integrating sensors into the structures in a way that Nondestructive Testing (NDT) is implemented continuously. One promising approach is to use Fiber Optic Sensors (FOS) to acquire strain signals, taking advantages of their capabilities over conventional sensors. Despite several works have developed Health and Usage Monitoring Systems (HUMS) using FOS for performing in-flight SHM in aircraft structures, automatic damage detection using the acquired signals has not been achieved in a robust way against environmental and operational variability, in all flight stages or considering different types of damages. In this work, a HUMS was developed and implemented in an Unmanned Aerial Vehicle (UAV) based on 20 Fiber Bragg Gratings (FBGs) embedded into the composite front spar of the aircraft's wing, a miniaturized data acquisition subsystem for gathering strain signals and a wireless transmission subsystem for remote sensing. The HUMS was tested in 16 flights, six of them were carried out with the pristine structure and the remaining after inducing different artificial damages. The in-flight data were used to validate a previously developed damage detection methodology based on strain field pattern recognition, or strain mapping, which utilizes machine learning algorithms, specifically a Self-Organizing Map (SOM)-based procedure for clustering operational conditions and Principal Component Analysis (PCA) in conjunction with damage indices for final classification. The performance of the damage detection demonstrated a highest accuracy of 0.981 and a highest F₁ score of 0.978. As a main contribution, this work implements in-flight strain monitoring, remote sensing and automatic damage detection in an operating composite aircraft structure.</p>

Topics

• impedance spectroscopy
• composite
• clustering
• machine learning